Explanation and discovery in aerodynamics

Gordon McCabe
December 22, 2005

Abstract
The purpose of this paper is to discuss and clarify the explanations
commonly cited for the aerodynamic lift generated by a wing, and to
then analyse, as a case study of engineering discovery, the aerodynamic
revolutions which have taken place within Formula 1 in the past 40 years.
The paper begins with an introduction that provides a succinct summary
of the mathematics of fluid mechanics.

1 Introduction
Aerodynamics is the branch of fluid mechanics which represents air flows and
the forces experienced by a solid body in motion with respect to an air mass. In
most applications of aerodynamics, Newtonian fluid mechanics is considered to
be empirically adequate, and, accordingly, a body of air is considered to satisfy
either the Euler equations or the Navier-Stokes equations of Newtonian fluid
mechanics. Let us therefore begin with a succinct exposition of these equations
and the mathematical concepts of fluid mechanics.
The Euler equations are used to represent a so-called ‘perfect’ fluid, a fluid
which is idealised to be free from internal friction, and which does not experience
friction at the boundaries which are defined for it by solid bodies.1 A perfect
Newtonian fluid is considered to occupy a region of Newtonian space Ω ⊂ R3 ,
and its behaviour is characterised by a time-dependent velocity vector field Ut ,
a time-dependent pressure scalar field pt , and a time-dependent mass density
scalar field ρt . The internal forces in any continuous medium, elastic or fluid,
are represented by a symmetric contravariant 2nd-rank tensor field σ, called
the Cauchy stress tensor, and a perfect fluid is one for which the Cauchy stress
tensor has the form σ = −pg, where g is the metric tensor representing the
spatial geometry. In component terms, with the standard Cartesian coordinates
(x1 , x2 , x3 ) on Euclidean space, σ ij = −pg ij = −pδ ij . Where ∇ is the covariant
derivative of the Levi-Civita connection corresponding to the Euclidean metric
tensor field on Newtonian space, the Euler equations are as follows:
· ¸
∂Ut
ρt + ∇Ut Ut = −grad pt + ft .
∂t
1 An ‘ideal’ fluid is the same thing as a ‘perfect’ fluid.

1
The scalar field ft represents the force exerted on the fluid, per unit volume,
by any external force field, such as gravity.2 Note that the gradient of pt is a
contravariant vector field, not to be confused with a covariant derivative of pt ,
something which would add a covariant index.
The reader will find in many texts that the Euler equations are expressed
without explicit use of the time subscript, and with (U · ∇)U or U · ∇U sub-
stituted in place of ∇U U . In terms of Cartesian coordinates, one can think of
U · ∇ as the differential operator obtained by taking the inner product of the
vector U with the differential operator ∂/∂x1 + ∂/∂x2 + ∂/∂x3 . In component
terms, one can write:

∂U i ∂U i ∂U i
((U · ∇)U )i = U 1 + U2 + U3 .
∂x1 ∂x2 ∂x3
The streamlines of a fluid at time t are the integral curves of the vector field
Ut .3 Given that Ut is capable of being time-dependent, the streamlines are also
capable of being time-dependent. The trajectory of a particle over time can be
represented by a curve in the fluid, φ : I → Ω, which is such that its tangent
vector φ0 (t) equals the fluid vector field U(φ(t),t) at position φ(t) at time t. The
trajectory of the particle which is at x ∈ Ω at time t = 0 can be denoted as
φx (t) = φ(x, t). The trajectories φx (t) followed by particles in the fluid only
correspond to streamlines in the special case of steady-state fluid flow, where
∂Ut /∂t = 0. The mapping φ : Ω×I → Ω enables one to define a time-dependent
set of mappings φt : Ω → Ω. Given any volume W at time 0, Wt = φt (W ) is the
volume occupied at time t by the collection of particles which occupied W at
time 0. φt is called the fluid flow map, and a fluid is defined to be incompressible
if the volume of any region W ⊂ Ω is constant in time under the fluid flow map,
(Chorin and Marsden 1997, p10). A fluid is defined to be homogeneous if ρ is
constant in space, hence a fluid which is both homogeneous and incompressible
must be such that ρ is constant in space and time.
The Euler equations are derived from the application of Newton’s second law
to a continuous medium, i.e., the force upon a fluid volume must equal the rate
of momentum transfer through that volume. From the conservation of mass,
one can derive the so-called continuity equation:
∂ρt
= div (ρt Ut ) .
∂t
In the case of an incompressible fluid, ∂ρt /∂t = 0 and div (ρt Ut ) = ρ div Ut ,
and, assuming ρ > 0, the continuity equation reduces to:

div Ut = 0 .
2 Given that a fluid possesses a mass density field ρ, it will also be self-gravitating, in

accordance with the Poisson equation: ∇2 Φ = −4πGρ. Needless to say, the strength of a
fluid’s own gravity is negligible in most applications of aerodynamics.
3 An integral curve α : I → Ω of a vector field V is such that its tangent vector α0 (s) equals

the vector field Vα(s) for all s ∈ I, where I is an open interval of the real line.

2
 In the more general case of a non-perfect fluid, the Cauchy stress tensor
takes the form (Chorin and Marsden 1997, p33):

σ = λ(div U )I + 2µD ,
where µ and λ are constants, and D is the ‘deformation’ tensor, or ‘rate of strain’
tensor, the symmetrisation of the covariant derivative of the velocity vector
field, D = Sym (∇U ). In the case of Cartesian coordinates in Euclidean space,
the components of the covariant derivative correspond to partial derivatives,
so U;ji = ∂U i /∂xj , and the components of the deformation tensor are Dij =
1/2(∂U i /∂xj + ∂U j /∂xi ).
The relevant equations for a fluid with such a Cauchy stress tensor, derived
again from Newton’s second law, are the Navier-Stokes equations. In the case
of an incompressible homogeneous fluid,4 the Navier-Stokes equations take the
following form:
· ¸
∂Ut 2
ρt − µ∇ Ut + ∇Ut Ut = −grad pt + ft .
∂t
The constant µ is the coefficient of viscosity of the fluid in question, and ∇2 is
the Laplacian, also represented as ∆ in many texts. With respect to a Cartesian
coordinate system in Euclidean space, each component U i of the velocity vector
field is required to satisfy the equation

· 2 i ¸
∂U i ∂ U ∂2U i ∂2U i ∂U i ∂U i ∂U i 1 ∂p f
−µ 2 + 2 + 2 + U1 + U2 + U3 =− + .
∂t ∂x1 ∂x2 ∂x3 ∂x1 ∂x2 ∂x3 ρ ∂xi ρ

The mass density ρ has been shifted here from the left-hand side to the right-
hand side of the equation, and the time subscript has been removed to avoid
clutter.
Only solutions of the Navier-Stokes equations are capable of representing
the presence of a boundary layer in the fluid flow adjacent to a solid body, a
phenomenon of crucial importance in aerodynamics. The only internal forces
in a perfect fluid are pressure forces, and these act normally to any surface in
the fluid. In contrast, the boundary layer is a consequence of friction and shear
stresses within a fluid, and such forces act tangentially to the surfaces which
separate adjacent layers in a fluid. Due to friction between a fluid and the
surface of a solid body, and the presence of shear stresses within the fluid, there
is a region of fluid adjacent to the solid body in which the velocity is less than
the free-flow velocity, and this region is the boundary layer. The fluid velocity
in the boundary layer tends towards zero approaching the surface of the solid
body. Fluids with different coefficients of viscosity µ have boundary layers of
different thickness, and different velocity gradients within the boundary layer.
4 Incompressible flow is a good approximation for aerodynamics at Mach numbers M < 0.4,
kU k
where M = c , the ratio of the flow speed to the speed of sound cs .
s

3
 Note that the Navier-Stokes equations, and fluid mechanics in toto, merely
provide a phenomenological approximation; a fluid is not a continuous medium,
but a very large collection of discrete molecules, interacting with each other.
Hence, fluid mechanics is not a fundamental physical theory. As a consequence,
there is no sense in which a body of air exactly realises a solution of the Navier-
Stokes equations. However, despite the fact that the Navier-Stokes equations
are only considered to be phenomenological, there are notable controversies and
difficulties in using these equations to explain the most important aerodynamic
phenomenon of all: the lift generated by a wing.

2 Bernoulli vs. Coanda

The simplest and most popular explanations of aerodynamic lift invoke the
Bernoulli principle, which, in turn, is derived from Bernoulli’s theorem. The
general version of Bernoulli’s theorem states that the following quantity,
1 p
k U k2 +w + + φ ,
2 ρ
is constant along the streamlines of a stationary, compressible, perfect isentropic
fluid, assuming p is a function of ρ, (Abraham, Marsden and Ratiu (1988),
p593). The gravitational potential for an external gravitational field is denoted
here as φ. The scalar field w is the internal energy density per unit mass of
the fluid,5 and an isentropic fluid is one for which the internal energy density w
depends only upon the mass density ρ. i.e., if the fluid undergoes compression,
the internal energy per unit mass may increase. Internal energy density is to be
distinguished from the kinetic energy density of the fluid, 1/2 ρ k U k2 . If the
internal energy density depends only upon the mass density, then this rules out
a change of internal energy density due to heat transfer between the fluid and
its environment, or a change of internal energy density due to the generation of
friction.
If one makes the same assumptions, but with the exception that the fluid is
both incompressible, and homogeneous throughout space, then the fluid will be
a constant density ρ0 throughout space and time; the internal energy density
will be constant w0 = w(ρ0 ) also; and the following quantity will be constant
along fluid streamlines:
1 p
k U k2 + + φ .
2 ρ0
The general version of Bernoulli’s theorem can be derived from the Euler equa-
tions,
· ¸
∂U
ρ + ∇U U = −grad p + f .
∂t
5 This
is related to the thermodynamic quantity of enthalpy per unit mass ² by the equation
² = w + p/ρ.

4
To provide a simple derivation, let us assume the case of an incompressible and
homogeneous fluid. Amongst other things, this entails that ∂U/∂t = 0, hence
Euler’s equations reduce to:

ρ0 (∇U U ) = −grad p + f .
We shall additionally assume that the fluid is irrotational. The covariant deriv-
ative can be expressed as follows,

k U k2
∇U U = grad + (curl U ) ∧ U ,
2
and an irrotational fluid is such that curl U = 0, hence Euler’s equations further
reduces to

k U k2
ρ0 (grad ) = −grad p + f .
2
Dividing both sides by ρ0 one obtains

k U k2 p f
grad = −grad + ,
2 ρ0 ρ0
and thence
µ ¶
k U k2 p f
grad + = ,
2 ρ0 ρ0
and
µ ¶
k U k2 p f
grad + +κ = ,
2 ρ0 ρ0
for any constant κ. Whilst f specifies the external force per unit volume, the
ratio f /ρ0 specifies the external force per unit mass, which is related to an
external potential φ by f /ρ0 = −grad φ. For any constant c it follows that
f
−grad (φ + c) = .
ρ0
Hence,
µ ¶
k U k2 p
grad + +κ = −grad (φ + c) .
2 ρ0
From this it follows that

k U k2 p
+ + κ = −φ − c ,
2 ρ0
and

k U k2 p
+ + φ = −(κ + c) ,
2 ρ0

5
which is our desired result because −(κ + c) is a constant.
From Bernoulli’s theorem, then, comes Bernoulli’s principle, which states
that an increase in fluid velocity corresponds to a decrease in pressure, and
a decrease in fluid velocity corresponds to an increase in pressure. Whilst
Bernoulli’s principle is often invoked to explain the phenomenon of aerody-
namic lift generated by the air flow around a wing profile, there are alternative
explanations which employ, in some combination: the ‘Coanda effect’, the no-
tion of circulation, and Newton’s third law. These alternative explanations
are, at the very least, equally legitimate to the Bernoulli-principle explanation,
and, amongst aerodynamicists, are considered to be superior to the Bernoulli-
explanation. There is also a long-standing popular misconception associated
with the Bernoulli-principle explanation, which has been widely disseminated,
but which is completely false.
The Bernoulli-principle approach explains the lift generated by a wing as
the consequence of lower pressure above the wing than below, resulting in a net
upward force upon the wing. The lower pressure above the wing corresponds
to faster air flow above the wing, in accordance with the Bernoulli principle.
The Bernoulli principle itself merely states that lower pressure corresponds to
faster airflow, and vice versa; it does not state that faster airflow causes lower
pressure, any more than lower pressure causes faster airflow. Faster airflow
does not have causal priority over lower pressure. However, at this point, the
misconceived popular explanation implicitly assigns causal priority to faster
airflow, explains the lower pressure as a consequence of the faster airflow, and
explains the faster airflow above the wing as a consequence of the fact that
wings have greater curvature above than below, and the air flowing over the top
therefore follows a longer path than the air flowing below. Conjoined with this
is a curious argument, which invokes the notion of a ‘packet’ of air, and claims
that a packet of air which is divided at the leading edge of a wing must rejoin
at the trailing edge, hence the air traversing the longer path over the top must
travel faster to rejoin its counterpart at the trailing edge. This ‘path-length’
explanation of lift does not require the air at the trailing edge of the wing to
be deflected downwards, despite the downdraught which can be experienced
directly beneath a passing aircraft. If this explanation were true, and the cause
of lift was merely the asymmetrical profiles of the upper and lower surfaces of
a wing, then it would not be possible for an aeroplane to fly upside-down. In
addition, experiment demonstrates that air passing over the upper surface of a
wing travels at such speed that packets of air divided at the leading edge of the
wing fail to rejoin at the trailing edge.
One alternative explanation argues that wings generate lift because they
deflect air downwards, and, by Newton’s third law, an action causes an equal
and opposite reaction, hence the wing is forced upwards. According to this
explanation, the trailing edge of a wing must point diagonally downwards to
generate lift, and this is achieved either by tilting the wing downwards with
respect to the flow of air, or by making the wing cambered, or both. Air is
deflected downwards by both the lower and upper surfaces of the wing. The
lower surface deflects air downwards in a straightforward fashion, given that the

6
wing is either tilted with respect to the direction of airflow, or the lower surface
is of a concave shape. However, the majority of the lift is generated by the
downwards deflection of the air which flows above the wing. This explanation
then depends upon the Coanda effect, the tendency of a stream of fluid to follow
the contours of a convex surface rather than continue moving in a straight line.
The flow over the surface of a wing is said to remain ‘attached’ to the wing
surface. An aircraft wing is said to ‘stall’ when the boundary layer on the upper
surface ‘detaches’ or ‘separates’, and is no longer guided downwards by the
contours of the upper surface. In this circumstance, the lifting force generated
by the upper surface of the wing suddenly becomes very small, and the lifting
force that remains is generally insufficient to support the weight of the aircraft.
To create lift at low speed, an aircraft must increase the angle of attack6 of the
wing, and the upper surface is carefully contoured to prevent flow detachment
under these conditions.
As stated in Section 1, whilst the region of fluid flow outside a boundary layer
can be idealized as a perfect fluid and represented by a solution of the Euler
equations, representation of the boundary layer itself requires the Navier-Stokes
equations. In particular, the inverse relationship between velocity and pressure,
encapsulated in Bernoulli’s principle, is only valid outside the boundary layer.
Thus, despite the velocity gradient across the boundary layer, normal to the
surface of the solid body, there is no corresponding pressure gradient. The
pressure across the boundary layer, normal to the surface of the solid body, is
represented to be almost equal to the pressure just outside the boundary layer.
The pressure inside the boundary layer changes along its length as the pressure
outside the layer changes. Towards the trailing edge of a wing, the pressure
outside the boundary layer builds, and the velocity outside the layer decreases,
approaching zero towards the aft ‘stagnation point’. This has the consequence
that the range of fluid flow velocities inside the boundary layer also decreases.
If the pressure increase towards the stagnation point is not sufficiently gradual,
then the boundary layer can separate from the surface of the wing before the
trailing edge. This occurs because the velocities inside the boundary layer are
already smaller than those outside, hence they can reach zero before the aft
stagnation point. After this, the velocities inside the boundary layer can reverse
direction, and this causes the flow to separate from the wing. Irrespective of
where the boundary layer separates, it then breaks up into eddies and forms a
turbulent wake behind the wing.
The boundary layer airflow which remains attached to the upper surface of
the wing, does so only because the pressure outside the boundary layer is slightly
higher than the pressure inside the boundary layer, so there is a pressure gradient
which forces the boundary layer to apparently adhere to the convex upper surface
of the wing. There is no genuine force of attraction between the wing surface
and the boundary layer airflow.
The lift generated by a wing is also often explained to be a consequence
6 The angle of attack is the angle subtended from the direction of airflow by the straight line

which joins the leading edge and trailing edge of the wing. This line is called the chordline,
and its length is called the chord of the wing.

7
of non-zero circulation in the airflow around the wing. Given a closed loop C
parameterized by s, in a fluid with a velocity vector field U , the circulation
around C is defined by the line integral
I
ΓC = hU, d/dsi ds ,
C
where d/ds is the tangent vector to the loop C. To understand how lift can
be explained by non-zero circulation, consider the idealized situation where the
airflow around a wing is a superposition of a uniform ‘freestream’ flow from left
to right, where the streamlines are parallel straight lines, and a pure circulatory
flow, where the streamlines are clockwise concentric circles. Taking the sum
of the velocity vector fields for the uniform and circulatory flow at each point,
the clockwise circulation is added to the freestream velocity above the wing,
and subtracted from it below. The Bernoulli principle can then be invoked
to explain the pressure differential above and below the wing, which, in turn,
explains the net upward force. The presence of circulation also accords with
the use of Newton’s third law to explain the uplift, because the circulatory flow
adds a downward component to the airflow in the wake of the wing, and this
corresponds to the downward deflection of air.
The notion of circulation can be used to quantitatively explain the lifting
force on a wing by the Kutta-Joukowski theorem. If C is a loop enclosing a
wing profile, and if an incompressible airflow of density ρ has a velocity vector
field which tends towards a constant U as the distance from the wing tends to
infinity, then the Kutta-Joukowski theorem asserts that the lifting force per unit
span7 on a notional wing of infinite span, is given by the equation

L = −ρ ΓC k U k n ,
where n is a unit vector orthogonal to U , (Chorin and Marsden 1997, p53). In
reality, the lifting force decreases towards the tips of a wing, hence the simpli-
fying assumption here of an infinite wing span, which allows one to treat the
flow as if it were two-dimensional. The unit vector n is orthogonal to U in a
two-dimensional plane which represents a longitudinal cross-section of the flow
around the wing.
The presence of a circular component to the airflow around a wing profile
cannot be explained without introducing the notions of viscosity, the bound-
ary layer, the Coanda effect, and boundary layer separation. The absence of
circulation requires the boundary layer to detach forward of the trailing edge
upon the upper surface of the wing. Upon the commencement of airflow, such
premature detachment does indeed occur, but this is a transient and unstable
situation, which, assuming left to right airflow, generates anti-clockwise vortices,
called starting vortices, until the boundary layer separation point migrates to
the trailing edge of the wing. Once the airflow above and below the wing is
detaching from the trailing edge, which is directed diagonally downward, there
is a non-zero clockwise circulatory component to the airflow.
7 The span is the width of a wing in a direction transverse to the airflow.

8
 The highest airflow velocity and lowest pressure occurs towards the front
of the upper surface of a wing. The presence of circulation raises the fluid
velocity above the wing, and the low radius of curvature at the front of the
wing accentuates the effect. This is because the velocity of fluid in circulation
is function of the radius of curvature. With the exception of very low radii,
the velocity of circulating fluid is inversely proportional to the radius. Hence,
where the radius of curvature of a wing is lowest, around the nose of the wing,
the fluid flow is accelerated to its highest velocity, and the pressure of the fluid
drops to its minimum. This effect is accentuated above the wing because of the
direction of the circulation. The radius of curvature increases towards the rear
of the wing, hence the velocity decreases and the pressure increases towards the
trailing edge of the upper surface.
Both the Bernoulli-principle explanation and the circulation explanation of
lift fall under the aegis of the deductive-nomological (DN) account of scien-
tific explanation. In such explanations, one explains certain phenomenal facts
by logically deriving them from the conjunction of general laws and particular
specified circumstances. As Torretti emphasises (1990, p22-24), such explana-
tions in physics typically require a re-conceptualization of the phenomenal facts
in the same terms in which the laws and circumstances are stated. A fluid
mechanics explanation of the lift generated by a wing represents the wing as
a compact subset M of Euclidean space, which is surrounded by a continu-
ous medium that possesses a fluid flow vector field, a pressure scalar field, and
a mass density scalar field satisfying the Navier-Stokes equations, and repre-
sents the presence of lift by a non-zero vertical component to the vector which
represents the net force upon the wing. The Bernoulli-principle
R explanation
represents the net force upon the wing as the integral − ∂M pn of the pressure
force over the surface ∂M of the wing, where n is an outward-pointing unit
normal; the circulation explanation
H represents the vertical force per unit span
as −ρ ΓC k U k n = −ρ [ C hU, d/dsi ds] k U k n, where C is a loop enclosing
the longitudinal profile of M , and n is a unit vector orthogonal to U .8
Although the Bernoulli principle can be used to determine the net force and
the lifting force upon a solid body once the pressure and velocity distribution
around the body have been specified, this flow regime is a consequence of viscos-
ity, and cannot be explained without recourse to the Coanda effect, circulation,
and the curvature of the wing. Hence the Bernoulli principle explanation is
subservient to an explanation of lift which employs the latter concepts.
Whilst the debate over the appropriate explanation for aerodynamic lift is a
debate over the appropriate fluid mechanical explanation for lift, fluid mechanics
is, of course, a continuum idealization, and, in reality, aerodynamic lift is the
consequence of the aggregate effect of the interactions between the molecules
in the air flow and the molecules in the surface of the wing. Charles Crummer
explains the Coanda effect in these terms as follows:
“In a steady flow over a surface, stream particles have only thermal velocity
8 Here we have neglected the fact that the lifting force per unit span decreases close to the
wing tips.

9
components normal to the surface. If the surface is flat, the particles that collide
with boundary layer particles are as likely to knock them out of the boundary
layer as to knock others in, i.e. the boundary layer population is not changed
and the pressure on the surface is the same as if there were no flow. If, however,
the surface is curved in a convex shape, the particles in the flow will tend to take
directions tangent to the surface, i.e. away from the surface, obeying Newtons
first law. As these particles flow away from the surface, their collisions with
the boundary layer thermal particles tend to “blow” those particles away from
the surface. What this means is that if all impact parameters are equally likely,
there are more ways a collision can result in a depletion of the boundary layer
than an increase in the boundary layer population. The boundary layer will
tend to increase in thickness and to depopulate; the pressure will reduce there.
Those particles in the flow that do interact with the stagnant boundary layer
will give some of their energy to particles there. As they are deflected back into
the flow by collisions with boundary layer particles, they are, in turn, struck by
faster particles in the flow and struck at positive impact parameters. They are
thus forced back toward the surface. This is how the flow is “attracted” to the
surface, the Coanda effect,” (Crummer 2005, p8).

3 Case Study: Aerodynamics in Formula 1

Formula 1 is an economic, technological and sporting activity, in which an ex-
tended physical object, the car, is equipped with an internal combustion engine,
capable of converting the chemical potential energy from a supply of fossil fuel
into useful work. A Formula 1 car is almost always in contact with a ground
plane, and this distinguishes the nature of the airflow from that experienced by
an aircraft in free flight.
As Peter Wright points out, “A Formula 1 car can be described aerody-
namically as a very low aspect ratio9 (0.38) bluff body in close proximity to
the ground (gap/chord = 0.005). It is surrounded by low aspect ratio rotat-
ing cylinders in contact with the ground, and it is equipped with leading and
trailing edge airfoils, the leading edge device also being in close proximity to
the ground. There are major internal flows, with heat addition (water, oil, and
brake cooling), an air intake, hot gas injection, and a large open cavity (the
cockpit). The flow is three-dimensional and has extensive areas of separated
flow and vorticity,” (2001, p125).
The objective of a racing car designer is to create a car which, under the
control of a human being, traverses a range of closed circuits at the greatest
average speed possible. This requires a car which accelerates, brakes, and cor-
ners as quickly as possible, within the constraints established by the technical
and sporting regulations. In physical terms, one wishes to maximize the lon-
gitudinal and lateral acceleration which the car is capable of generating. The
tyres of a car provide the contact surfaces between the car and the road surface,
9 The aspect ratio for a wing of span s and area S is s2 /S; for a rectangular wing, the

aspect ratio is the ratio of the span to the chord, (Katz 1995, p115).

10
and the frictional grip provided by the tyres determines the upper limit on the
longitudinal and lateral forces which the car is capable of generating.
The grip generated by a tyre is a function of the force pushing that tyre
onto the road surface. The maximum horizontal force Fh which a tyre can
generate is Cf Fv , where Cf is the tyre’s peak coefficient of friction, and Fv is
the peak downward vertical force on the tyre. As a consequence, the greater the
weight of a car, the greater the grip generated by its tyres. However, greater
weight generally fails to increase the longitudinal and lateral acceleration of a car
because the increase in the longitudinal and lateral force which the car is capable
of generating is cancelled out by the greater inertia which is a concomitant of
the greater weight.
Two aerodynamic revolutions can be identified in the history of Formula 1:
the discovery of ‘downforce’ by means of inverted wing profiles, and the dis-
covery of ground effect downforce. As a consequence of the first revolution,
Formula 1 designers are able to generate an aerodynamic force which is directed
downwards (‘downforce’) by attaching appendages to the car which are, in ef-
fect, inverted versions of the wing profiles on aircraft. This increases the force
pushing the tyres onto the road surface without any increase in inertia. Al-
though downforce brings an inevitable drag force penalty with it, the increase
in the lateral accelerative capabilities of a Formula 1 car increases the average
speed of a Formula 1 car over most types of closed circuit. Both the drag force
and the downforce are proportional to the square of the velocity of a car. The
drag force is given by
1
Fdrag ≈ CD Aρv 2 ,
2
where CD is the drag coefficient, a dimensionless number determined by the
exact shape of the car and its angle of attack; A is the frontal area; ρ is the
density of air; and v is the speed of the car. The downforce is given by
1
Fdownf orce ≈ CL Aρv 2 ,
2
where CL is the coefficient of lift, again determined by the exact shape of the car
and its angle of attack. In the case of a modern Formula 1 car, the lift-to-drag
ratio CL /CD has a typical value of, say, 2.5 (Wright 2001, p125), so downforce
dominates performance.
Inverted wing profiles were introduced to Formula 1 rather belatedly, after
a process of gradual discovery occurring over many decades, and involving mul-
tiple individuals within the community of automotive engineers, (Hughes 2005,
p119-121; Katz 1995, p245-250). In fact, the idea of aerodynamic downforce has
a provenance which can be traced back to 1928. Inverted wing profiles were ini-
tially used to provide stability or to counteract the tendency a car might have
to lift at high speeds. In 1928, Opel’s experimental rocket-powered cars, the
RAK 1 and RAK 2, were equipped with inverted wings on either side between
the wheels to counteract high-speed lift (see Figures 1 and 2). In 1956 a Swiss
engineer and amateur racing driver called Michael May experimented with an

11
 Figure 1: Opel’s 1928 experimental rocket car, the RAK 1, with side-wings.

Figure 2: Opel’s RAK 2, with enlarged side-wings.

inverted wing mounted over the cockpit of his Porsche 550 Spyder (see Figure
3). Michael and his brother Pierre had recalled the use of such wings upon the
Opel RAK. When May’s car proved faster than the works Porsches, Porsche
lobbied successfully for the appendage to be banned, under the pretext that it
obscured the vision of following drivers, and May failed to pursue the idea any
further.
Perhaps the most influential innovator in the field of racing car aerodynamics
was Texan oil magnate, engineer and driver Jim Hall. Hall’s Chaparral company
linked up with the Chevrolet R&D department, headed by Frank Winchell, and
competed in US sportscar championships such as Can-Am in the 1960s. In 1961,
the Chaparral 1 sports car experienced lift at high speed, and Bill Mitchell, chief
stylist of General Motors in the 1950s and 1960s, suggested using an inverted
wing (Wright 2001, p299). In 1963 protruding front-mounted wings were fitted
to prevent the front wheels of the Chaparral 2 from lifting off the ground (Wright
2001, p300). In 1965, the Chaparral 2C was fitted with a rear wing mounted

12
 Figure 3: Michael May’s 1956 Porsche 550 Spyder, with inverted wing.

on pivots, with a driver-adjustable angle of attack, and in 1966 the concept was
extended to a dramatic high wing on the Chaparral 2E, (see Figure 4).
In 1966 the McLaren F1 team tested wings with great success, but due to
lack of resources, the team never found the time to revisit the idea, as recounted
by chief designer Robin Herd (Hughes 2005, p120): “We didn’t want anyone else
to twig, so we took the wings off, quietly put them in the back of the truck and
continued with our normal testing. We decided we would look into it further,
in private, when we had the time. But an F1 team in those days was so madly
understaffed that we never got round to looking at it properly.”

Figure 4: The high-winged Chaparral 2E of 1966.

13
 In November 1967, Jim Clark raced an American Indycar called a Vollst-
edt, which possesed small wings, and during the Tasman winter series Clark
recounted the grip and stability produced by this car to one of his Lotus me-
chanics. The mechanic extemporised a make-do wing from a helicopter rotor,
which was fitted to Clark’s car, and swiftly removed, but not before a Ferrari
engineer had taken photographs of it. At the 1968 Belgian Grand Prix, Fer-
rari appeared with full inverted rear wings (see Figure 5), and Brabham did
likewise on the day after Ferrari’s wings first appeared. As Mark Hughes re-
counts, “Ferrari’s chief engineer Mauro Forghieri, his memory perhaps triggered
by that Tasman photo taken by one of his staff of that experimental Lotus
wing, had recalled that Michael May - the engineer with whom he had worked
perfecting Ferrari’s fuel injection system a few years earlier - had once made a
wing for that sports car of his. ‘Michael was a friend as well as a consultant,’
says Forghieri. ‘He told me about the improvement in handling of his winged
Porsche. The Chaparral convinced us even more about the idea’,” (Hughes
2005, p121). Given that May’s idea was, in turn, inspired by the Opel RAK,
the introduction of downforce to Formula 1 in 1968 can be traced back to Opel’s
experimental rocket car of 1928!

Figure 5: The winged Ferrari 312 of 1968.

The utility of downforce seems so obvious in retrospect, that its rather be-
lated application in Formula 1 seems puzzling. The stories told about scien-
tific discovery often seem to involve either a deliberate, conscious, trend of
thought that finally reaches some goal, or sudden leaps of the imagination, or
serendipitous discoveries made by minds receptive to new ideas. In the case

14
of aerodynamic downforce, an example of engineering discovery, none of these
descriptions seem to apply. Seeking to explain this, Peter Wright asserts that
“the possible magnitude of the downforce that could be generated was not ap-
preciated, despite all the data being available within the aeronautical industry.
It was probably also considered that the tires (sic) would not be able to sus-
tain the additional work and that the drag penalty associated with generating
the downforce would result in a net loss of average speed...It may also have
been simply an attitude among designers, most of whom came from produc-
tion car backgrounds-the cars did not have wings. It took the free spirits of
Frank Winchell and Jim Hall, and their measurement and simulation groups at
Chevrolet R&D and Chaparral, to cast aside convention and explore the pos-
sibilities of this new and now obvious way to increase performance,” (Wright
2001, p124). The reluctant adoption of downforce in Formula 1 seems to be an
example of the trammelling effect of convention in academic and research com-
munities. It is a demonstration that the significance of new empirical facts often
go unrecognised when they are not subsumed within the appropriate conceptual
scheme: aerodynamics were only considered to be relevant to Formula 1 to the
extent that they enabled a designer to minimise drag. One might say that a
particular engineering paradigm existed in Formula 1, unquestioned, until the
late 1960s, when the amount of suggestive empirical data built to such an extent
that a revolution was unavoidable.
The second revolution in Formula 1 aerodynamics occurred approximately
a decade after the first, with the introduction of the Lotus T78 in 1977, and its
further development, the Lotus T79 in 1978 (see Figure 6). It was discovered
that large amounts of downforce could be generated from the airflow between
the underbody of the car and the ground plane. In particular, low pressure
could be created underneath the car by using the ground plane almost like the
floor of a venturi duct. The ceiling of these venturi ducts took the form of
inverted wing profiles mounted in sidepods between the wheels of the car. The
decreasing cross-sectional area in the throat of these ducts, and the inverted
wing profile accelerated the airflow and created low pressure in accordance with
the Bernoulli principle. The gap between the bottom of the sidepods and the
ground was sealed by so-called ‘skirts’. When the rules permitted it, the skirts
were suspended from the sidepods with a vertical degree of freedom (Katz 1995,
p200) to maintain a constant seal with the ground under changes in the attitude
and ride height of the car.
Brabham responded to the Lotus ‘wing car’ with the BT46B ‘fan car’, in
which low pressure was created under the car by a rear-mounted fan driven off
the gearbox. Such a car is able to generate downforce independently of the
speed of airflow under the car, and the BT46B won its debut race in 1978, and
was promptly banned by the governing body.
Whilst the inverted wing profiles in the sidepods of ground effect cars were
often dubbed ‘sidewings’, and the bottom of the car was often dubbed the
‘underwing’, these names are potentially misleading because the downforce gen-
erated is largely dependent upon the presence of the ground plane, and would
not be generated in free flow. Note also that the tunnels created under the car

15
 Figure 6: The skirted ground effect Lotus T79 of 1978.

are not strictly venturi ducts either. In a venturi duct, the sides of the duct do
not move with respect to each other. In contrast, underneath such a car the
ground plane and the airflow are both moving with respect to the walls and
ceiling of the ducts.
Again, this second revolution in Formula 1 aerodynamics could have been
implemented many years before it was. Jim Hall and Chevrolet R&D had tried
shaping the entire underbody as an inverted wing profile in the 1960s, and the
1970 Chaparral 2J had been equipped with two rear-mounted fans to lower the
air pressure under the car, and flexible plastic skirts to seal the low pressure
area. The Chaparral 2J was banned by the governing body of Can-Am at the
end of the season under lobbying from fellow competitors. In 1969, Formula 1
engineer Peter Wright designed an inverted wing body shape for BRM which
yielded promising results in the Imperial College wind-tunnel, (Wright 2001,
p301). However, the design never saw the light of day due to a management
upheaval. Peter Wright joined Lotus some years later, and made a serendipitous
discovery in the Imperial College wind-tunnel in 1975: “By the end of a week of
tunnel testing, the strong wooden model would have been so modified with card,
modeling clay, and sticky tape...that it had usually lost most of its structural
integrity. Toward the end of one of the weeks in the tunnel, I noticed that it was
becoming almost impossible to obtain consistent balance readings. Something
was wrong. Looking carefully at the model, it became clear that the side pods
were sagging under load and that as the speed of the tunnel increased, they

16
sagged even more. That indicated two things: (1) that the side pods had
started to generate downforce, and (2) that it had something to do with the gap
between their edges and the ground.
“Thin wire supports restored the side pods to their correct position and
stopped them from sagging - no downforce and consistent balance readings.
Next we taped card skirts to seal the gap between the edge of the side pods and
the ground, leaving only approximately 1mm (0.04in) gap. The total downforce
on the car doubled for only a small increase in drag!” (ibid., p302).
From 1983 onwards, the technical regulations in Formula 1 required the cars
to have flat bottoms with no skirts, and from mid-1994 onwards, the cars were
required to have flat bottoms with a central step running the length of the
underbody, yet substantial ground effect downforce could still be generated by
accelerating the airflow under the car. A flat plane inclined downwards with
respect to the airflow will generate downforce, and this effect will be amplified
by ground effect. In addition, more downforce can be generated by using an
upswept ‘diffuser’ between the wheels at the rear of the car to force air upwards
by means of the Coanda effect.
The Coanda effect is also utilised on a modern Formula 1 car with the pur-
pose, not of generating downforce directly, but of guiding and conditioning air-
flow in one place, as a means of maximising downforce elsewhere. For example,
the rear of a modern Formula 1 car is tightly tapered between the rear wheels,
rather like the neck and shoulders of a coke-bottle. By means of the Coanda
effect, the air flowing along the flanks of the sidepods adheres to the waisted
contours at the rear, and the airflow here is duly accelerated, creating lower
pressure. In itself, this tranverse pressure differential on either side of the car
cancels out, and creates no net force. However, the accelerated airflow between
the rear wheels and over the top of the diffuser does raise the velocity of the
air exiting the diffuser. The accelerated airflow over the lower surface of the
rear wing also contributes to raising the flow rate through the diffuser, thereby
enhancing the downforce it generates. In addition, coaxing air away from the
rear tyres, and injecting high speed airflow into the region of flow separation
behind the car, both contribute to reducing drag.
The Coanda effect is also used by the so-called ‘bargeboards’, aerodynamic
appendages typically sited between the trailing edge of the front wheels and the
leading edge of the sidepods (see Figure 7). Bargeboards are used to guide tur-
bulent air from the front wing wake, away from the vital airflow underneath the
car. In addition, the lower trailing edge of a bargeboard creates a vortex which
travels down the outer lower edge of the sidepod, acting as a surrogate skirt,
helping to seal the lower pressure area under the car. Such techniques, and the
continued utility of ground effect in Formula 1, demonstrate the irreversibility
of conceptual revolutions, even in the case of an engineering activity which faces
increasingly constrained technical regulations.

17
 Figure 7: Modern Formula 1 bargeboard.

References
[1] Abraham, R., Marsden, J.E., and Ratiu, T. (1988). Manifolds, Tensor
Analysis, and Applications, Second Edition. New York: Springer-Verlag.
[2] Chorin, A.J. and Marsden, J.E. (1997). A Mathematical Introduction to
Fluid Mechanics, Third Edition. New York: Springer Verlag.
[3] Crummer, C.A. (1998). Aerodynamics at the particle level,
arXiv:nlin.CD/0507032.
[4] Hughes, M. (2005). Speed Addicts. London: CollinsWillow.

[5] Katz, J. (1995). Race Car Aerodynamics. Cambridge, MA: Bentley pub-
lishers.

[6] Torretti, R. (1990). Creative Understanding. London: University of Chicago

Press.

[7] Wright, P. (2001). Formula 1 Technology. Warrendale, PA: Society of Au-

tomotive Engineers.

18